24小时热门版块排行榜

返回列表

jianqin

金虫 (正式写手)

应助: 7 (幼儿园)
金币: 3649.9
红花: 2
帖子: 627
在线: 394小时
虫号: 863555
注册: 2009-10-05
专业: 质谱分析

[求助] 关于GAMESS中QMTTOL参数的设置问题

QMTTOL，linear dependence threshhold，根据GAMESS手册说明，一般默认设为1.0E-6，如果使用了弥散函数，是多环分子，需要调整到3.0E-6.

当我用b3lyp 6-31g+(d,p)做几何优化时，用了默认值1.0E-6，输出文件有warning提示：
POSSIBLE LINEAR DEPENDENCE PROBLEMS DETECTED, INPUT QMTTOL= 1.0E-06, WITH  2 OVERLAP EIGENVALUES LYING BETWEEN 1.0E-05 AND QMTTOL.
IF YOU EXPERIENCE SCF CONVERGENCE DIFFICULTIES WITH THE DENSITY ALMOST BUT NOT QUITE CONVERGED, YOU MAY NEED TO LOOSEN -QMTTOL- BY A FACTOR 5 TO 10.

结果没有收敛，我讲QMTTOL值改为5.0E-6，再做计算，输出文件还是有warning提示，还多了一项：
...WARNING (SYMORB.SRC)...
THE INPUT BASIS SET CONTAINS APPROXIMATE LINEAR DEPENDENCE.
THE SMALLEST EIGENVALUE OF THE OVERLAP MATRIX IS 1.476998E-06
THERE ARE 2 EIGENVALUES LESS THAN QMTTOL=  5.00E-06
EIGENVECTORS BELOW -QMTTOL- ARE DROPPED FROM THE MO SPACE,
IN ORDER TO ELIMINATE THE APPROXIMATE LINEAR DEPENDENCE.

THIS WILL MOST LIKELY CHANGE THE ENERGY ON THE ORDER OF TENS OF MICROHARTREES,
DEPENDING ON THE NUMBER OF ORBITALS REMOVED.  THE VALUE OF QMTTOL CAN BE
CHANGED IN $CONTRL: JOBS WITH EIGENVALUES BELOW 1.0E-07 PROBABLY WON'T
CONVERGE.  EIGENVALUES BETWEEN 1.0E-07 AND 1.0E-06 MAY REQUIRE LOOSENING OF
THE DENSITY CRITERION (CONV IN $SCF), MORE ACCURATE INTEGRAL EVALUATION
(ICUT=11 IN $CONTRL), MORE ACCURATE DIRECT SCF FOCK MATRIX
CONSTRUCTION (FDIFF=.FALSE. IN $SCF), OR CHANGING THE SCF CONVERGER
(DIIS=.TRUE. IN $SCF).

                  * * * * * WARNING * * * * *
POSSIBLE LINEAR DEPENDENCE PROBLEMS DETECTED, INPUT QMTTOL= 5.0E-06,
WITH    2 OVERLAP EIGENVALUES LYING BETWEEN 1.0E-05 AND QMTTOL.
IF YOU EXPERIENCE SCF CONVERGENCE DIFFICULTIES WITH THE DENSITY ALMOST BUT
NOT QUITE CONVERGED, YOU MAY NEED TO LOOSEN -QMTTOL- BY A FACTOR 5 TO 10.

结果还是不收敛，请问这是什么原因？这个QMTTOL参数的意义到底是什么？如何影响最后的计算结果？

回复此楼